Prediction system, prediction method, and program

Abstract

Provided are a prediction system, a prediction method, and a program that can more easily predict the characteristics of a fiber-reinforced resin. A prediction system 1 comprises an acquiring unit (CPU 110) that acquires, with respect to a fiber-reinforced resin, first measurement information measured by a first measuring device, and second measurement information measured by a second measuring device different from the first measuring device and / or thickness information of the fiber-reinforced resin measured by a third measuring device, and a predicting unit (CPU 110) that predicts a characteristic within the fiber-reinforced resin on the basis of the acquired first measurement information and the second measurement information and / or the thickness information, wherein: the first measuring device is any one of a millimeter-wave measuring device, an ultrasonic measuring device, an X-ray diffraction device, and an X-ray Talbot-Lau device; and the second measurement information includes information derived from a chemical structure.

Description

Prediction system, prediction method, and program

[0001] The present invention relates to a prediction system, a prediction method, and a program.

[0002] In recent years, fiber-reinforced resins have been attracting attention in various fields, such as space, aircraft, automobiles, ships, fishing rods, electrical components, electronic components, home appliance components, parabolic antennas, bathtubs, flooring materials, and roofing materials. Under these circumstances, there is a demand for accurate prediction of the properties of fiber-reinforced resins. In particular, for fiber-reinforced resins reinforced with short fibers, there is a demand for accurate prediction of the fiber orientation within the fiber-reinforced resin. X-ray computed tomography (CT) as described in Patent Document 1 is known as a means for determining fiber orientation.

[0003] Japanese Patent Application Laid-Open No. 2020-126023

[0004] However, the X-ray CT disclosed in Patent Document 1 has problems such as restrictions on the size of the fiber-reinforced resin, a limited range of analysis, and time-consuming measurement and analysis.

[0005] An object of the present invention is to provide a prediction system, a prediction method, and a program that can more easily predict the properties of fiber-reinforced resin.

[0006] In order to solve the above problem, the prediction system according to the present invention comprises: an acquisition unit that acquires first measurement information measured by a first measurement device for a fiber-reinforced resin, second measurement information measured by a second measurement device different from the first measurement device, and / or thickness information of the fiber-reinforced resin measured by a third measurement device; and a prediction unit that predicts characteristics within the fiber-reinforced resin based on the acquired first measurement information, the second measurement information, and / or the thickness information, wherein the first measurement device is any one of a millimeter-wave measurement device, an ultrasonic measurement device, an X-ray diffraction device, and an X-ray Talbot-Lau device, and the second measurement information includes information derived from a chemical structure.

[0007] In order to solve the above-mentioned problems, the prediction method according to the present invention includes: an acquisition step of acquiring, for a fiber-reinforced resin, first measurement information measured by a first measuring device, second measurement information measured by a second measuring device different from the first measuring device, and / or thickness information of the fiber-reinforced resin measured by a third measuring device; and a prediction step of predicting properties within the fiber-reinforced resin based on the acquired first measurement information, the second measurement information, and / or the thickness information, wherein the first measuring device is a millimeter-wave measuring device, an ultrasonic measuring device, an X-ray diffraction device, or an X-ray Talbot-Lau device, and the second measurement information includes information derived from a chemical structure.

[0008] In order to solve the above problem, the program of the present invention causes a computer of a prediction device to function as: an acquisition unit that acquires first measurement information measured by a first measurement device for a fiber-reinforced resin, second measurement information measured by a second measurement device different from the first measurement device, and / or thickness information of the fiber-reinforced resin measured by a third measurement device; and a prediction unit that predicts characteristics within the fiber-reinforced resin based on the acquired first measurement information, second measurement information, and / or thickness information, wherein the first measurement device is a millimeter-wave measurement device, an ultrasonic measurement device, an X-ray diffraction device, or an X-ray Talbot-Lau device, and the second measurement information includes information derived from a chemical structure.

[0009] According to the present invention, the properties of fiber reinforced resin can be predicted more easily.

[0010] 1 is a diagram showing the overall configuration of a prediction system. FIG. 2 is a block diagram showing a schematic configuration of a prediction device. FIG. 3 is a diagram explaining the principle of a Talbot interferometer. FIG. 4 is a flowchart showing a prediction process. FIG. 5 is a flowchart showing a machine learning method for a trained model. FIG. 6 is an image of a flat test piece. FIG. 7 is an image of measurement points on a flat test piece. FIG. 8 is an image of measurement points on a flat test piece. FIG. 9 is an image of measurement points on a flat test piece. FIG. 10 is an image of a flat test piece. FIG. 11 is an image of a box-shaped test piece. FIG. 12 is an image of a test piece cut out from a flat test piece. FIG. 13 is an image of a test piece cut out from a box-shaped sample. FIG. 14 is an image of a flat test piece. FIG. 15 is an image of a box-shaped test piece. FIG. 16 is an image of measurement points on a box-shaped test piece.

[0011] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings, but the scope of the invention is not limited to the illustrated examples.

[0012] <Configuration of Prediction System 1> Fig. 1 is a diagram showing the overall configuration of the prediction system 1. As shown in Fig. 1, the prediction system 1 includes, for example, a prediction device 100, a first measurement device 200, a second measurement device 300, and a third measurement device 400. The prediction device 100, the first measurement device 200, the second measurement device 300, and the third measurement device 400 are connected to a communication network N. Note that Fig. 1 shows a case where there is one second measurement device 300, but there may be multiple second measurement devices 300. Furthermore, the third measurement device 400 may not be present. The communication network N may be a LAN (Local Area Network), a WAN (Wide Area Network), the Internet, or the like.

[0013] The prediction system 1 predicts the properties of fiber-reinforced resin, such as fiber orientation, tensile strength, and warpage. Fiber-reinforced resin is a composite material containing a filler and a resin.

[0014] Examples of the resin include known thermosetting resins and thermoplastic resins. Specific examples include polyolefin resins such as polyethylene resin (PE), polypropylene resin (PP), and maleic anhydride-modified polypropylene (MAHPP), epoxy resins, phenolic resins, unsaturated polyester resins, vinyl ester resins, polycarbonate resins, polyester resins, polyamide (PA) resins, liquid crystal polymer resins, polyethersulfone resins, polyetheretherketone resins, polyarylate resins, polyphenylene ether resins, polyphenylene sulfide (PPS) resins, polyacetal resins, polysulfone resins, polyimide resins, polyetherimide resins, polystyrene resins, modified polystyrene resins, AS resins (copolymers of acrylonitrile and styrene), ABS resins (copolymers of acrylonitrile, butadiene, and styrene), modified ABS resins, MBS resins (copolymers of methyl methacrylate, butadiene, and styrene), modified MBS resins, polymethyl methacrylate (PMMA) resins, and modified polymethyl methacrylate resins. The resin contained in the composite material may be one of these resins, or a mixture of two or more of these resins.

[0015] Fillers are added to resins, for example, to improve the strength of composite materials. Fillers are added to resins at a volumetric concentration of 0.1% to 50%. Fillers may be, for example, fibrous or particulate. Examples of fibrous fillers include glass fiber (GF), carbon fiber (CF), aramid fiber, alumina fiber, silicon carbide fiber, boron fiber, and silicon carbide fiber. Examples of CF that can be used include polyacrylonitrile (PAN), pitch, cellulose, and hydrocarbon vapor-grown carbon fiber and graphite fiber. Examples of GF that can be used include E-glass and S-glass. It is preferable that the composite material contain at least one of glass fiber (GF) and carbon fiber (CF).

[0016] The particulate filler is, for example, calcium carbonate (CaCo ３ ), talc (Mg３ Si ４ O １０ (OH) ２ ), barium sulfate (BaSO ４ ), mica (Si, Al, Mg, K), aluminum hydroxide (Al(OH) ３ ), magnesium hydroxide (Mg(OH) ２ ), titanium oxide (TiO ２ ), zinc oxide (ZnO ２ ), antimony oxide (Sb ２ O ３ ), kaolin clay (Al ２ O ３ 2SiO ２ ・2H ２ The filler contained in the composite resin material may be one of these, or a mixture of two or more of these.

[0017] The composite material may contain a sensitivity adjuster. A sensitivity adjuster is a sample that functions like a contrast agent used in X-ray imaging, enabling measurement of the composite material with greater accuracy and sensitivity. Measuring a composite material containing a sensitivity adjuster using the second measurement device 300 enables measurement with greater accuracy. For example, when the second measurement device 300 is a Raman spectrometer, using zirconium tungstate as the sensitivity adjuster changes the Raman shift, allowing measurement information regarding the optical properties of the composite material to be generated with greater accuracy. For example, when the second measurement device 300 is a fluorescence microscope, using a fluorescent dye as the sensitivity adjuster enables measurement information of the composite material to be generated with greater accuracy.

[0018] It is preferable that the sensitivity adjuster contained in the composite material has little effect on the physical properties of the composite material. This allows the composite material measured by the second measurement device 300 to be used in a molded product. A test piece of the composite material containing the sensitivity adjuster may be prepared for measurement by the second measurement device 300.

[0019] (Prediction device 100) The prediction device 100 is a computer such as a PC, a smartphone, a tablet terminal, etc. The prediction device 100 is configured to be connectable to a first measurement device 200, a second measurement device 300, and a third measurement device 400, and transmits and receives various information to and from each device.

[0020] FIG. 2 is a block diagram showing a schematic configuration of the prediction device 100.

[0021] 2, the prediction device 100 includes a CPU (Central Processing Unit) 110, a ROM (Read Only Memory) 120, a RAM (Random Access Memory) 130, storage 140, a communication interface 150, a display unit 160, and an operation reception unit 170. Each component is connected to each other via a bus so as to be able to communicate with each other.

[0022] The CPU 110 controls the above components and performs various arithmetic processing in accordance with programs recorded in the ROM 120 and the storage 140. The CPU 110 functions as an acquisition unit that acquires first measurement information measured by a first measurement device and second measurement information measured by a second measurement device different from the first measurement device for the fiber reinforced resin. The CPU 110 functions as a prediction unit that predicts the properties within the fiber reinforced resin based on the acquired first measurement information and second measurement information. The functions of the acquisition unit and prediction unit will be described in detail below.

[0023] The ROM 120 stores various programs and various data.

[0024] The RAM 130 serves as a working area for temporarily storing programs and data.

[0025] The storage 140 stores various programs including an operating system and various data. For example, an application for predicting the properties of a fiber-reinforced resin using a trained classifier is installed in the storage 140. The storage 140 may also store first measurement information and second measurement information (described below) acquired from the first measuring device 200, the second measuring device 300, and the third measuring device 400. The storage 140 may also store trained models used as classifiers and training data used in machine learning.

[0026] The communication interface 150 is an interface for communicating with other devices. A wired or wireless communication interface conforming to various standards is used as the communication interface 150. The communication interface 150 is used, for example, to receive the first measurement information and the second measurement information from the first measurement device 200, the second measurement device 300, and the third measurement device 400, and to transmit the prediction results to a server or the like for storage.

[0027] The display unit 160 includes an LCD (liquid crystal display), an organic EL display, etc., and displays various information. The display unit 160 may be configured with viewer software, a printer, etc.

[0028] The operation reception unit 170 includes a touch sensor, a pointing device such as a mouse, a keyboard, etc., and receives various operations from the user. The display unit 160 and the operation reception unit 170 may form a touch panel by superimposing a touch sensor serving as the operation reception unit 170 on the display surface serving as the display unit 160.

[0029] (First measuring device 200) The first measuring device 200 measures the fiber-reinforced resin and generates first measurement information regarding the orientation of the fibers. The first measuring device 200 is any one of a millimeter-wave measuring device, an ultrasonic measuring device, an X-ray diffraction (XRD) device, and an X-ray Talbot-Lau device.

[0030] The millimeter-wave measurement device applies millimeter waves to the fiber-reinforced resin and measures the fiber-reinforced resin's response to the millimeter waves. The ultrasonic measurement device applies ultrasonic waves to the fiber-reinforced resin and measures the fiber-reinforced resin's response to the ultrasonic waves, i.e., acoustic properties. The X-ray diffraction device irradiates the fiber-reinforced resin with X-rays to measure the radiation properties of the composite material. The X-ray Talbot-Lau device measures the fiber-reinforced resin and generates Talbot information, described below, as first measurement information.

[0031] Here, an X-ray Talbot-Lau device will be described. (Photography Using a Talbot Interferometer and a Talbot-Lau Interferometer) Here, a photographing method using a Talbot interferometer and a Talbot-Lau interferometer will be described. As shown in FIG. 3 , when X-rays emitted from a radiation source 11 pass through a first grating 14, the transmitted X-rays form images at regular intervals in the z direction. These images are called self-images, and the phenomenon of forming self-images is called the Talbot effect. A second grating 15 is placed roughly parallel to the self-image at the position where the self-images are formed, and a Moiré fringe image (indicated by Mo in FIG. 3 ) is obtained by the X-rays that pass through the second grating 15. That is, the first grating 14 forms a periodic pattern, and the second grating 15 converts the periodic pattern into Moiré fringes. If a fiber-reinforced resin is present between the radiation source 11 and the first grating 14, the phase of the X-rays is shifted by the fiber-reinforced resin, and the Moiré fringes on the Moiré fringe image are distorted at the edges of the fiber-reinforced resin, as shown in FIG. 3 . This disturbance in the moiré fringes can be detected by processing the moiré fringe image, and an image of the fiber-reinforced resin can be produced. This is the principle of the Talbot interferometer.

[0032] A multi-slit 12 is placed between the radiation source 11 and the first grating 14 and close to the radiation source 11, and X-ray imaging is performed using a Talbot-Lau interferometer. The Talbot interferometer is based on the premise that the radiation source 11 is an ideal point radiation source, but in actual imaging, a focal spot with a relatively large focal spot diameter is used, and therefore the multi-slit 12 produces an effect as if X-rays were being irradiated from a series of multiple point radiation sources. This is the X-ray imaging method using a Talbot-Lau interferometer, and even when the focal spot diameter is relatively large, it is possible to produce the same Talbot effect as with a Talbot interferometer.

[0033] With this type of X-ray Talbot-Lau device, at least three types of images (two-dimensional images) can be reconstructed (referred to as reconstructed images) by capturing a moiré image Mo ( FIG. 3 ) of a fiber-reinforced resin using a method based on the principles of fringe scanning and analyzing the moiré image Mo using Fourier transform. These three types of images are: an absorption image (same as a normal X-ray absorption image) that visualizes the average component of the moiré fringes in the moiré image Mo; a differential phase image that visualizes the phase information of the moiré fringes; and a small-angle scattering image that visualizes the visibility of the moiré fringes. It is also possible to generate even more types of images by, for example, recombining these three types of reconstructed images.

[0034] Next, the three or more small-angle scattering images for each of the prepared relative angles are aligned. Because the sample is rotated, each image is returned to the specified angle.

[0035] Finally, fitting is performed for each pixel with a sine wave, and fitting parameters are extracted. A sine wave graph is a graph in which the horizontal axis represents the relative angle between the sample and the lattice, and the vertical axis represents the small-angle scattering signal value of a certain pixel. The amplitude, average, and phase of the sine wave are obtained as fitting parameters. An image showing the amplitude value for each pixel is called an "amp image," an image showing the average value for each pixel is called an "ave image," and an image showing the phase for each pixel is called a "pha image." The amp image, ave image, and pha image are collectively called "orientation images." The fitting method is not limited to sine waves; for example, the angle (phase) with the greatest intensity can be calculated as θ ０ An ellipse with the highest intensity as a and the lowest intensity as b may be fitted to the following equation (1) expressed in polar coordinates with the position r(θ). In this case, in accordance with the names used in sine wave fitting, an image with a value (a-b) / 2 corresponding to the amplitude of each pixel is called an "amp image," an image showing a value (a+b) / 2 corresponding to the average value of each pixel is called an "ave image," and an image showing a value (a+b) / 2 corresponding to the θ of each pixel is called an "amp image." ０ The image showing the major axis a, minor axis b, and phase θ of the signal intensity for each pixel may be referred to as a "pha image." ０ may be assigned to the image as an orientation image.

[0036] Such an X-ray Talbot-Lau device generates a Talbot image containing information about the filler orientation of the fiber-reinforced resin. That is, the Talbot information generated by the X-ray Talbot-Lau device is information about the Talbot image. The Talbot image is an image generated by the X-ray Talbot-Lau device photographing the fiber-reinforced resin and using the Talbot effect. Images that have been subjected to image processing such as the orientation image are also included in the Talbot image. A reconstructed image reconstructed from the moiré image Mo is also included in the Talbot image.

[0037] (Second Measuring Device 300) The second measuring device 300 is a device for measuring a fiber-reinforced resin and generating information derived from the chemical structure included in the second measurement information. Here, the information derived from the chemical structure is the chemical properties and physical properties of the fiber-reinforced resin, such as acoustic properties, atomic properties, electrical properties, magnetic properties, mechanical properties, optical properties, radiation properties, thermal properties, and surface condition. The second measuring device 300 is preferably a device capable of measuring the fiber-reinforced resin non-destructively. This allows the fiber-reinforced resin measured by the second measuring device 300 to be used in subsequent manufacturing processes, etc.

[0038] The second measuring device 300 is, for example, an infrared spectrometer, a Raman spectrometer, a fluorescence spectrophotometer, an ultraviolet-visible spectrophotometer, or a nuclear magnetic resonance (NMR) device.

[0039] Infrared spectrometers and Raman spectrometers irradiate fiber-reinforced resins with electromagnetic waves to measure the response of composite materials to electromagnetic waves, i.e., their optical properties. Fluorescence spectrophotometers irradiate fiber-reinforced resins with light and measure the fluorescence emitted from the fiber-reinforced resin. UV-visible spectrophotometers irradiate fiber-reinforced resins with light and measure the ultraviolet and visible light transmitted through the fiber-reinforced resin and / or the ultraviolet and visible light reflected from the fiber-reinforced resin. Nuclear magnetic resonance spectrometers are devices that can analyze the molecular structure and physical properties of fiber-reinforced resins. Fiber-reinforced resins can be measured in their solid form, or they can be dissolved in a solvent and the extract measured. Nuclear magnetic resonance spectrometers can obtain information not only on the structure of fiber-reinforced resins, but also on intermolecular and intramolecular interactions, molecular mobility, and other information.

[0040] (Third Measuring Device 400) The third measuring device 400 is a device for measuring the fiber-reinforced resin and generating shape information (such as thickness information) that is the thickness of the fiber-reinforced resin included in the second measurement information. The third measuring device 400 is, for example, a micrometer, a serration, or a probe-type, laser-type, or image-type three-dimensional measuring machine.

[0041] <Functions of Prediction Device 100> The CPU 110 of the prediction device 100 functions as an acquisition unit, an extraction unit, a prediction unit, and a display control unit by reading a program stored in the storage 140 and executing the process.

[0042] The acquisition unit acquires the first measurement information generated by the first measurement device 200 and the second measurement information generated by the second measurement device 300 and the third measurement device 400. The first measurement information and the second measurement information acquired by the acquisition unit preferably include information about the same region of the composite resin. This makes it possible to predict the characteristics of a specific region of the composite resin with high accuracy.

[0043] The extractor extracts feature quantities from each of the first measurement information and the second measurement information acquired by the acquirer. The feature quantities are extracted from, for example, a spectrum, an image, or the like acquired by the acquirer.

[0044] For example, since the first measurement information is the millimeter wave spectrum of the fiber reinforced resin, the extractor extracts frequency characteristics, main components, etc. When the first measurement information is information related to an ultrasound image of the fiber reinforced resin, the extractor extracts from the first measurement information the reflection intensity of the front and back surfaces of the fiber reinforced resin, the time it takes for the ultrasound to pass through the fiber reinforced resin, main components, etc.

[0045] When the first measurement information is information relating to the X-ray diffraction spectrum of a fiber-reinforced resin, the extraction unit extracts the diffraction intensity at a specific angle, the angle of the diffraction maximum, the main component, the crystallinity, etc. from the first measurement information.

[0046] The feature quantity extracted from the Talbot information, which is the first measurement information, may be the Talbot image itself, such as an orientation image (amp image, ave image, pha image), or may be an image signal value acquired from a specific region of the Talbot image. The feature quantity extracted from the Talbot information may be the degree of orientation, the orientation angle, etc.

[0047] Furthermore, the feature extracted from the Talbot information may be the eccentricity ecc. The eccentricity ecc can be calculated, for example, by calculating σ1=ave+amp (corresponding to the maximum value of the small-angle signal value) and σ2=ave-amp (corresponding to the minimum value of the small-angle signal value) using the signal values ​​amp and ave obtained from the orientation image, using the following equation (2). The orientation image may include an image (ecc image) showing the eccentricity ecc for each pixel.

[0048]

[0049] For example, when the second measurement information is information about the infrared absorption spectrum of the fiber-reinforced resin, the extraction unit extracts from the second measurement information the absorption intensity at a specific wavenumber, the integrated value of the absorption intensity, the wavelength showing the absorption maximum, the main component, etc. When the second measurement information is, for example, a Raman spectrum of the fiber-reinforced resin, the extraction unit extracts from the second measurement information the emission intensity at a specific wavelength, the integrated value of the emission intensity, the wavelength showing the emission maximum, the main component, etc. When the second measurement information is, for example, an emission spectrum obtained from a fluorescence spectrophotometer of the fiber-reinforced resin, the extraction unit extracts the emission intensity at a specific wavelength, the excitation wavelength or emission wavelength showing the emission maximum, the main component, etc. When the second measurement information is, for example, an absorption spectrum or reflection spectrum of the fiber-reinforced resin obtained from an ultraviolet-visible spectrophotometer, the extraction unit extracts the absorption intensity or reflection intensity at a specific wavelength, the wavelength showing the absorption maximum or the wavelength showing the reflection maximum, the main component, etc. When the second measurement information is, for example, a nuclear magnetic resonance spectrum obtained from a nuclear magnetic resonance apparatus for a fiber-reinforced resin, the chemical shift value of the peak, the integrated value of the peak, the main component, etc. are extracted.

[0050] The extraction unit may extract a plurality of feature quantities from each of the first measurement information and the second measurement information. Furthermore, the feature quantities extracted from the second measurement information may be principal components obtained by principal component analysis of the spectrum. Furthermore, the feature quantities extracted from the second measurement information may be principal components obtained by principal component analysis of the measured waveform.

[0051] The acquisition unit may acquire information from which feature quantities have been extracted. That is, the first measurement information and the second measurement information may be information from which feature quantities have been extracted from information about the fiber-reinforced resin measured by the first measuring device 200 and the second measuring device 300.

[0052] The prediction unit predicts the properties of the fiber-reinforced resin based on the first measurement information and the second measurement information acquired by the acquisition unit. Specifically, the prediction unit predicts the properties of the fiber-reinforced resin using a trained classifier with the feature quantities of the first measurement information and the second measurement information extracted by the extraction unit as input.

[0053] The display control unit causes the display unit 160 to output information relating to the properties of the fiber reinforced resin predicted by the prediction unit.

[0054] The processing executed by the prediction device 100 will be described in detail below.

[0055] <Prediction Processing> Fig. 4 is a flowchart showing the procedure of the prediction processing executed in the prediction device 100. The processing of the prediction device 100 shown in the flowchart in Fig. 4 is stored as a program in the storage 140 of the prediction device 100, and is executed by the CPU 110 controlling each unit.

[0056] (Steps S101, S102) The prediction device 100 first acquires first measurement information obtained by measuring the fiber reinforced resin using the first measuring device 200 (step S101). Next, the prediction device 100 acquires second measurement information obtained by measuring the fiber reinforced resin using the second measuring device 300 and the third measuring device 400, and thickness information of the fiber reinforced resin (step S102). The order in which the first measurement information, the second measurement information, and the thickness information are acquired does not matter, and they may be acquired simultaneously.

[0057] The prediction device 100, for example, acquires first measurement information from the first measuring device 200, and second measurement information and fiber-reinforced resin thickness information from the second measuring device 300 and the third measuring device 400. The first measuring device 200, the second measuring device 300, and the third measuring device 400 may store the first measurement information, the second measurement information, and the fiber-reinforced resin thickness information in another device such as a server, and the prediction device 100 may acquire the first measurement information, the second measurement information, and the fiber-reinforced resin thickness information from the other devices.

[0058] (Step S103) The prediction device 100 extracts feature quantities from each of the first measurement information, the second measurement information, and the thickness information of the fiber reinforced resin obtained in the processes of steps S101 and S102.

[0059] (Step S104) The prediction device 100 inputs the feature quantities of the first measurement information, the second measurement information, and the fiber-reinforced resin thickness information extracted in the process of step S103 into a classifier that has been previously trained by machine learning to predict the properties of the fiber-reinforced resin. For example, the classifier is trained by machine learning using a learning method described below using teacher data that includes the feature quantities of the first measurement information, the second measurement information, and the fiber-reinforced resin thickness information of a large number of fiber-reinforced resins prepared in advance, as well as measured values ​​of the properties of each of the fiber-reinforced resins. Specifically, the classifier is trained by machine learning using the feature quantities of the first measurement information, the second measurement information, and the fiber-reinforced resin thickness information of each of the fiber-reinforced resins as input data and the measured values ​​of the properties of each of the fiber-reinforced resins as output data. As a result, the prediction device 100 can predict the properties of the fiber-reinforced resin by inputting the feature quantities extracted from the first measurement information, the second measurement information, and the fiber-reinforced resin thickness information into the classifier. If the property of the fiber-reinforced resin is fiber orientation, the measured values ​​of the property of the fiber-reinforced resin are obtained, for example, by X-ray CT. When the characteristic of the fiber reinforced resin is tensile strength, the measurement value is obtained using, for example, a Tensilon universal testing machine RTI-1225 (manufactured by A&D Co., Ltd.) When the characteristic of the fiber reinforced resin is warpage, the measurement value is obtained using, for example, a CNC image measuring system NEXIV VMR-3020 (manufactured by Nikon Corporation).

[0060] The classifier may perform machine learning using first measurement information, second measurement information, and thickness information of the fiber-reinforced resins related to a plurality of fiber-reinforced resins as input data and measurement values ​​of the properties of each of the plurality of fiber-reinforced resins as output data. Furthermore, the information input to the classifier is not limited to the feature quantities of the first measurement information, the second measurement information, and the thickness information of the fiber-reinforced resins. For example, in addition to the feature quantities of the first measurement information, the second measurement information, and the thickness information of the fiber-reinforced resins, information at the time of manufacture may be input to the classifier and used as information for learning and prediction.

[0061] (Step S105) The prediction device 100 generates a prediction result of the properties of the fiber reinforced resin based on the output of the classifier in the process of step S104.

[0062] (Step S106) The prediction device 100 outputs the prediction result generated in the process of step S105. For example, the prediction device 100 displays the value of the property of the fiber reinforced resin predicted in the process of step S104 on the display unit 160 together with information about the fiber reinforced resin.

[0063] The information acquired in step S102 may be only the second measurement information or only the thickness information.

[0064] <Learning Process> Next, a machine learning method for a trained model used in a classifier will be described.

[0065] FIG. 4 is a flowchart showing a machine learning method for a trained model.

[0066] In the process of FIG. 4 , machine learning is performed using a large number (i sets) of data sets as training sample data. The data sets are inputs of feature quantities of the first measurement information, the second measurement information, and the thickness information of the fiber-reinforced resins prepared in advance, and outputs measured values ​​of the properties of each of the fiber-reinforced resins. The learning device (not shown) functioning as a classifier may be, for example, a standalone high-performance computer using a CPU and a GPU processor, or a cloud computer. Below, a learning method using a neural network formed by combining perceptrons such as deep learning in the learning device will be described. However, various other methods may be applied, including, for example, random forests, decision trees, support vector machines (SVMs), logistic regression, k-nearest neighbors, and topic models.

[0067] (Step S111) The learning device reads learning sample data, which is teacher data. If it is the first time, the first set of learning sample data is read, and if it is the i-th time, the i-th set of learning sample data is read.

[0068] (Step S112) The learning device inputs the input data from the read learning sample data to the neural network.

[0069] (Step S113) The learning device compares the prediction result of the neural network with the correct answer data.

[0070] (Step S114) The learning device adjusts the parameters based on the comparison results. For example, the learning device adjusts the parameters by performing a process based on back-propagation (back-propagation) so that the difference in the comparison results becomes smaller.

[0071] (Step S115) If the learning device has completed processing of all data from the first to i-th sets (YES), the process proceeds to step S116; if not (NO), the process returns to step S111, the next learning sample data is read, and the process from step S111 onwards is repeated.

[0072] (Step S116) The learning device determines whether or not to continue learning. If it decides to continue (YES), the process returns to step S111, and it executes the processes from the first set to the i-th set again in steps S111 to S115. If it decides not to continue (NO), the process proceeds to step S117.

[0073] (Step S117) The learning device stores the trained model constructed in the processing up to this point and ends the processing (END). The storage destination includes the internal memory of the prediction device 100. In the processing of FIG. 4 described above, the trained model generated in this manner is used to predict the properties of the fiber-reinforced resin.

[0074] [Effects] As described above, the prediction system 1 includes an acquisition unit (CPU 110) that acquires first measurement information measured by a first measuring device, second measurement information measured by a second measuring device different from the first measuring device, and / or thickness information of the fiber-reinforced resin measured by a third measuring device, and a prediction unit (CPU 110) that predicts properties within the fiber-reinforced resin based on the acquired first measurement information, the second measurement information, and / or the thickness information, where the first measuring device is any of a millimeter-wave measuring device, an ultrasonic measuring device, an X-ray diffraction device, and an X-ray Talbot-Lau device, and the second measurement information includes information derived from the chemical structure. Therefore, the properties of the fiber-reinforced resin can be predicted more easily without using X-ray CT.

[0075] The prediction method includes steps of acquiring first measurement information measured by a first measuring device, second measurement information measured by a second measuring device different from the first measuring device, and / or thickness information of the fiber-reinforced resin measured by a third measuring device (steps S101 and S102), and predicting properties within the fiber-reinforced resin based on the acquired first measurement information, the second measurement information, and / or the thickness information (step S104). The first measuring device is a millimeter-wave measuring device, an ultrasonic measuring device, an X-ray diffraction device, or an X-ray Talbot-Lau device, and the second measurement information includes information derived from the chemical structure. This makes it possible to predict the properties of the fiber-reinforced resin without using X-ray CT, making it easier to predict the properties of the fiber-reinforced resin.

[0076] The program also causes the computer of the prediction device 100 to function as an acquisition unit (CPU 110) that acquires first measurement information measured by a first measurement device, second measurement information measured by a second measurement device different from the first measurement device, and / or thickness information of the fiber-reinforced resin measured by a third measurement device, and a prediction unit (CPU 110) that predicts properties within the fiber-reinforced resin based on the acquired first measurement information, the second measurement information, and / or the thickness information, where the first measurement device is a millimeter-wave measurement device, an ultrasonic measurement device, an X-ray diffraction device, or an X-ray Talbot-Lau device, and the second measurement information includes information derived from the chemical structure. This makes it possible to predict the properties of fiber-reinforced resins without using X-ray CT, making it easier to predict the properties of fiber-reinforced resins.

[0077] The effects of the present invention will be explained using the following examples, although the technical scope of the present invention is not limited to the following examples.

[0078] Example 1 First, multiple resin molded product (fiber-reinforced resin) samples were prepared. These samples were prepared using the following combinations of three types of resin, three types of filler, and two filler concentrations (weight ratios). The resin and filler were mixed in the desired ratios in advance using a Labo Plastomill® extruder manufactured by Toyo Seiki Seisaku-sho, Ltd. Pellets were then prepared. The resin molded product samples were molded using an SE100D injection molding machine manufactured by Sumitomo Heavy Industries, Ltd. Two types of sample shapes were used: flat plate-shaped test pieces measuring 80 mm x 80 mm x 1 mm (t) and 80 mm x 80 mm x 5 mm (t), with one piece having a film gate G (Figure 6, Sample Image SP1; front view, Sample Image SP2; side view).

[0079] Resins: polypropylene (Noblen (registered trademark) W101 manufactured by Sumitomo Chemical Co., Ltd.), polyamide 66 (Leona 1300S manufactured by Asahi Kasei Corporation), polycarbonate (Iupilon (registered trademark) H-3000R manufactured by Mitsubishi Engineering Plastics Corporation); Fillers: PAN (polyacrylonitrile)-based carbon fibers (CF-N manufactured by Nippon Polymer Sangyo Co., Ltd.), glass fibers (CS3J-960 manufactured by Nitto Boseki Co., Ltd.); Filler concentrations: 10%, 30%.

[0080] Next, each of the 18 types of resin molded product samples was measured using the following measuring device. Measurements were taken at three locations (AR1, AR2, and AR3) on the sample image SP3 shown in Figure 7, and a total of 54 types of data were obtained.

[0081] Millimeter-wave measurement device: Measurements were performed using a device combining the following: Millimeter waves were generated using a PC-controlled Windfreak Technologies oscillator, which was multiplied by 4 using an AT Microweave Limited multiplier amplifier to output a frequency of 30 to 50 GHz. The sample was irradiated using a millimeter-wave antenna. The sample was placed on an aluminum plate with a 4 mm x 10 mm slit. The millimeter-wave intensity was detected using a detector from SAGE Millimeter Inc. Analog-to-digital conversion was then performed using an ADC, and data acquisition was performed using a PC. Measurements were performed by changing the orientation of the sample to 0 degrees, 45 degrees, 90 degrees, and 135 degrees, as shown in Figure 8. Figure 8 shows an image of millimeter waves passing through a slit irradiating a fiber-reinforced resin (sample image SP4) at various angles (irradiation ranges I1, I2, I3, I4). The signal strengths from 30 GHz to 50 GHz were averaged to determine the millimeter-wave intensity. The millimeter-wave intensity at 45 degrees, 90 degrees, and 135 degrees was normalized using the millimeter-wave intensity at 0 degrees to obtain the characteristic quantities.

[0082] Ultrasonic measurement device (ultrasonic imaging device FineSAT III (Hitachi)) The probe P was tilted as shown in Figure 9 and rotated to perform measurements from four directions. The probe P shown in Figure 9 is equipped with an ultrasonic output unit P1, and from the left in Figure 9, the orientation of the probe P was rotated 0°, 45°, 90°, and 135° (based on the downward direction (dotted line) on the page) relative to the fiber-reinforced resin (sample image SP5; front view, sample image SP6; side view, sample image SP7; bottom view). Using an ultrasonic device, measurements were performed using the surface echo / first bottom echo method specified in JIS Z2353. The surface gain, bottom gain, and surface gain / bottom gain were extracted from the data obtained by the measurement. The numerical values ​​of each angle were normalized to the numerical value of 0 degrees and used as feature quantities. The probe and sample were in contact with each other via an echo gel pad (Yasojima Proceed Co., Ltd.).

[0083] X-ray diffraction equipment (Smart Lab, manufactured by Rigaku Corporation) Measurements were performed by changing the orientation of the sample (Sample Image SP8) relative to the rotation direction of the XRD detector. Measurements were performed using the transmission method. Figure 10 shows the entire sample, but in reality, the sample was cut to a size that could fit into the sample holder before measurement. In Figure 10, area A1 indicates the area irradiated with X-rays, and area A2 indicates the area detected by the detector.

[0084] X-ray Talbot-Lau apparatus (apparatus described in JP 2019-184450 A) Measurements were performed by rotating the sample at angles of 0 degrees, 45 degrees, 90 degrees, and 135 degrees. The signal intensities of the small-angle scattering images obtained by imaging the visibility of the moiré fringes were averaged over the area of ​​interest to determine the signal intensity at each angle. The signal intensity at each angle was normalized by the signal intensity at 0 degrees to obtain a feature value.

[0085] Infrared spectrometer (FT-IR) (AVATAR370 manufactured by Thermo Fisher Scientific); Raman spectrometer (compact Raman spectrophotometer PR-1w (manufactured by JASCO Corporation)); spectrofluorometry: three-dimensional fluorescence spectrum of fluorescence spectrophotometer F-7000 (manufactured by Hitachi High-Tech); NMR: solid-state NMR JMN-ECA400W (JEOL); Thickness: micrometer OMC-150MX (Mitutoyo).

[0086] <Method of Obtaining Correct Values ​​(Creating a Trained Classifier)> Fiber orientation was measured using an X-ray CT scanner SKYSCAN 1272 (Bruker Japan Co., Ltd.). Samples were cut into approximately 5 mm x 10 mm pieces and measured. Fiber orientation was calculated using image analysis software. As shown in the fiber-reinforced resin shown in Figure 11 (sample image SP9; front view, sample image SP10; side view), the reference direction was the direction of resin flow during injection molding, the angle between the fiber and the reference direction was θ, and the average value of cos θ was taken as the fiber orientation. Note that this was the average value for 50 fibers. The classifier was trained using the above data.

[0087] <Verification using a trained learning machine> Next, verification was performed using a trained learning machine. Box-shaped test specimens measuring 80 mm x 80 mm x 30 mm were produced, as shown in Figure 12. The plate thickness of each part was 3 mm. The left image in Figure 12 (sample image SP11) is a top view of the box-shaped test specimen, and the right image in Figure 12 (sample image SP12) is a bottom view of the box-shaped test specimen. The reference direction AX1 is the direction of the arrow, and area AR4 is to be measured. The resin and filler type were the same as those used in the trained data, and test specimens with a filler concentration of 20% were produced. Six types in total.

[0088] Resin: Polypropylene (Noblen (registered trademark) W101 manufactured by Sumitomo Chemical Co., Ltd.), polyamide 66 (Leona 1300S manufactured by Asahi Kasei Corporation), polycarbonate (Iupilon (registered trademark) H-3000R manufactured by Mitsubishi Engineering Plastics Corporation); Filler: PAN (polyacrylonitrile)-based carbon fiber (CF-N manufactured by Nippon Polymer Sangyo Co., Ltd.), glass fiber (CS3J-960 manufactured by Nitto Boseki Co., Ltd.); Filler concentration: 20%;

[0089] Various measurements were performed in the same way as when acquiring the learning data. X-ray CT measurements were also performed as the correct values.

[0090] In each example, the classifier was trained using the data types shown in Table I. The trained classifier was used to input the analysis data of the verification data and determine the degree of orientation. The difference between the predicted value of the degree of orientation determined by the classifier and the measured value of the degree of orientation determined by X-ray CT was calculated as the error as follows:

[0091]

[0092] The errors in the table are expressed as relative values ​​to Comparative Example 101. This shows that the error is significantly reduced by combining two measured values. Furthermore, it can be seen that the error is reduced by combining three measured values.

[0093] <Example 2> Furthermore, the strength prediction was verified. First, the following data were used for learning. Explanatory variables: fiber orientation, fiber concentration (set value), fiber type, and thickness (design value) obtained in Example 1. Objective variables: tensile strength. The tensile strength was measured using a test piece cut out from a flat test piece (sample image SP13) as shown in Figure 13. The shape was that of JIS K7139 A12 (excluding thickness). The tensile strength was measured using a Tensilon universal testing machine RTI-1225 (A&D Co., Ltd.) under conditions of a jig distance of 50 mm and a tensile speed of 50 mm / sec.

[0094] The verification data was the tensile strength of a test piece cut out from the box-shaped sample. The cut-out position was as shown in Figure 14 (sample image SP14). Figure 14 is a view from the resin inlet side of gate G.

[0095] The results are summarized in Table II.

[0096]

[0097] Table II shows that the strength prediction error is significantly reduced by using a fiber orientation prediction that combines two measurements, and even more so by combining three measurements.

[0098] Example 3: Using the fiber orientation data obtained in Example 1, an analysis was performed with the amount of warpage (ease of warpage) as the objective variable. In Example 2, three dumbbell-shaped test pieces were cut out from the flat plate test piece and used to learn the objective variable, but in Example 3, the amount of warpage of the flat plate was obtained without cutting out any flat plate test pieces. Specifically, the amount of warpage of the flat plate was obtained using fiber orientation data from three locations as the explanatory variable.

[0099] First, the following data were used for learning. Explanatory variables: fiber orientation (fiber orientation at three locations on the same test piece) determined in Example 1, fiber concentration (set value), fiber type, and thickness (design value). Objective variables: amount of warpage.

[0100] The amount of warpage was measured after the sample was left for two weeks in a high-temperature, high-humidity chamber at 25°C and 50% RH. The amount of warpage was measured by determining the amount of displacement from the reference plane AR5, which was 10 mm from the gate side of the flat surface. When the amount of warpage varied across the width, the value at the point with the largest absolute value of change was used (Fig. 15, Sample Image SP15: Front View, Sample Image SP16: Side View). Measurements were performed using a CNC image measuring system NEXIV VMR-3020 (manufactured by Nikon Corporation).

[0101] For the verification data, a box-type test specimen was used as in Examples 1 and 2, but the fiber orientation was measured at points A4, A5, and A6 on the box-type test specimen (sample image SP17) shown in Figure 16. Furthermore, warpage was measured using the above-mentioned measurement method, as shown in Figure 17 (sample image SP18 in Figure 17; side view, sample image SP19 in Figure 17; bottom view). The results are summarized in Table III.

[0102] Table III shows that the warpage prediction error is significantly reduced by using a fiber orientation prediction that combines two measurements, and even more so by combining three measurements.

[0103] The above-described configurations of the prediction device 100 and the prediction system 1 are merely the main configurations described in order to explain the features of the above-described embodiments and examples, but are not limited to the above-described configurations and may be modified in various ways within the scope of the claims. Furthermore, configurations that are included in general prediction systems are not excluded.

[0104] For example, the prediction device 100 may include components other than the above-described components, or may not include some of the above-described components.

[0105] Furthermore, the prediction device 100, the first measurement device 200, the second measurement device 300, and the third measurement device 400 may each be configured by a plurality of devices, or may each be configured by a single device.

[0106] Furthermore, the functions of each component may be realized by other components. For example, the first measurement device 200, the second measurement device 300, and the third measurement device 400 may be integrated into the prediction device 100, and some or all of the functions of the first measurement device 200, the second measurement device 300, and the third measurement device 400 may be realized by the prediction device 100.

[0107] Furthermore, the processing units of the flowcharts in the above embodiments are divided according to the main processing content in order to facilitate understanding of each process. The classification of the processing steps does not limit the scope of the present invention. Each process can be divided into more processing steps. Furthermore, one processing step may execute more processes.

[0108] The means and methods for performing various processes in the systems according to the above-described embodiments can be realized by either dedicated hardware circuits or programmed computers. The programs may be provided, for example, on computer-readable recording media such as flexible disks and CD-ROMs, or online via a network such as the Internet. In this case, the programs recorded on the computer-readable recording media are typically transferred to and stored in a storage unit such as a hard disk. The programs may also be provided as standalone application software, or may be incorporated into the software of the device as a function of the system.

[0109] The present disclosure can be used in a prediction system, a prediction method, and a program.

[0110] REFERENCE SIGNS LIST 100 Prediction device 110 CPU (acquisition unit, prediction unit) 120 ROM 130 RAM 140 Storage 150 Communication interface 160 Display unit 170 Operation reception unit 200 First measurement device 300 Second measurement device 400 Third measurement device

Claims

1. A prediction system comprising: an acquisition unit that acquires first measurement information measured by a first measuring device for a fiber-reinforced resin, second measurement information measured by a second measuring device different from the first measuring device, and / or thickness information of the fiber-reinforced resin measured by a third measuring device; and a prediction unit that predicts characteristics within the fiber-reinforced resin based on the acquired first measurement information, the second measurement information, and / or the thickness information, wherein the first measuring device is any of a millimeter-wave measuring device, an ultrasonic measuring device, an X-ray diffraction device, and an X-ray Talbot-Lau device, and the second measurement information includes information derived from a chemical structure.

2. The prediction system according to claim 1, wherein the second measuring device is any one of an infrared spectrometer, a Raman spectrometer, a fluorescence spectrophotometer, an ultraviolet-visible spectrophotometer, and a nuclear magnetic resonance device.

3. A prediction system as described in claim 1, wherein the prediction unit predicts the characteristics using machine learning with the first measurement information and the second measurement information as explanatory variables.

4. A prediction system as described in claim 3, wherein in the machine learning, measurement information measured by X-ray CT is used as a target variable.

5. A prediction method comprising: an acquisition step of acquiring, for a fiber-reinforced resin, first measurement information measured by a first measuring device, second measurement information measured by a second measuring device different from the first measuring device, and / or thickness information of the fiber-reinforced resin measured by a third measuring device; and a prediction step of predicting characteristics within the fiber-reinforced resin based on the acquired first measurement information, the second measurement information, and / or the thickness information, wherein the first measuring device is a millimeter-wave measuring device, an ultrasonic measuring device, an X-ray diffraction device, or an X-ray Talbot-Lau device, and the second measurement information includes information derived from a chemical structure.

6. A program that causes a computer of a prediction device to function as: an acquisition unit that acquires first measurement information measured by a first measuring device for a fiber-reinforced resin, second measurement information measured by a second measuring device different from the first measuring device, and / or thickness information of the fiber-reinforced resin measured by a third measuring device; and a prediction unit that predicts characteristics within the fiber-reinforced resin based on the acquired first measurement information, the second measurement information, and / or the thickness information, wherein the first measuring device is a millimeter-wave measuring device, an ultrasonic measuring device, an X-ray diffraction device, or an X-ray Talbot-Lau device, and the second measurement information includes information derived from a chemical structure.

Images